Avoid cloning jump threading state when possible
The current implementation of jump threading passes most of its time cloning its state. This PR attempts to avoid such clones by special-casing the last predecessor when recursing through a terminator.
This is not optimal, but a first step while I refactor the state data structure to be sparse.
The two other commits are drive-by.
Fixes https://github.com/rust-lang/rust/issues/116721
r? `@oli-obk`
ignore `llvm::Lld` if lld is not enabled
People are having trouble ([ref. zulip thread](https://rust-lang.zulipchat.com/#narrow/stream/326414-t-infra.2Fbootstrap/topic/MSVC.20Runtime.20mismatch.20when.20building.20LLD)) when they don't want to build `lld` for their custom distribution tarballs even with `lld = false` in their config.toml. This is because it is not controlled by `lld_enabled` flag. This change ensures that `llvm:Lld` is controlled by lld configuration.
Additionally, `lld = true` is set by default for dist profile, because we have been building it all along and this maintains that behavior.
try-job: x86_64-mingw
Use proper `ImplTraits` in `insert_inference_vars_for_impl_trait`
Fixes#17199 and fixes#17403
In the previous implementation, I passed `rpits` as a function parameter and used `idx` of `ImplTraitId` for indexing `ImplTrait`.
4e836c622a/crates/hir-ty/src/infer.rs (L881-L887)
But that `idx` is rather a "local" one, so in the cases like mentioned issues, the async function that can be expanded roughly as
```rust
type TypeAlias = impl Something;
fn expanded_async() -> impl Future<Output = TypeAlias> { ... }
```
there are two bundles of `ImplTraits`; one for the `impl Future` and the other one for `TypeAlias`.
So using `idx` with `rpits` returns `ImplTrait` for `impl Future` even if we are asking for `TypeAlias` and this caused a stack overflow.
This PR is a fix for that implementation miss 😅
ast: Standardize visiting order
Order: ID, attributes, inner nodes in source order if possible, tokens, span.
Also always use exhaustive matching in visiting infra, and visit some discovered missing nodes.
Unlike https://github.com/rust-lang/rust/pull/125741 this shouldn't affect anything serious like `macro_rules` scopes.
rustdoc: use current stage if download-rustc enabled
When using download-rustc, using stage 1 rustdoc results in the wrong librustc_driver being used.
```sh
$ ./build/host/stage1/bin/rustdoc --version
./build/host/stage1/bin/rustdoc: error while loading shared libraries: librustc_driver-7ff02ed05016d515.so: cannot open shared object file: No such file or directory
```
This change fixes that by not cutting the stage if download-rustc is enabled.
Fixes for 32-bit SPARC on Linux
This PR fixes a number of issues which previously prevented `rustc` from being built
successfully for 32-bit SPARC using the `sparc-unknown-linux-gnu` triplet.
In particular, it adds linking against `libatomic` where necessary, uses portable `AtomicU64`
for `rustc_data_structures` and rewrites the spec for `sparc_unknown_linux_gnu` to use
`TargetOptions` and replaces the previously used `-mv8plus` with the more portable
`-mcpu=v9 -m32`.
To make `rustc` build successfully, support for 32-bit SPARC needs to be added to the `object`
crate as well as the `nix` crate which I will be sending out later as well.
r? nagisa
patch `rust-lld` and `ld.lld` on NixOS
When `rustc` uses its self-contained lld, we also need to patch `rust-lld` and `ld.lld`.
The `rpath` for `rust-lld` is `$ORIGIN/../../../:$ORIGIN/../lib`, so I use `--add-rpath` instead of `--set-rpath`, which should be easier to maintain.
I also changed `src/bootstrap/src/core/download.rs`, even this doesn't fix any known issues.
For the `lld-wrapper.sh` of lld, refer to: https://github.com/rust-lang/rustc-dev-guide/pull/1999.
Remove more `PtrToPtr` casts in GVN
This addresses two things I noticed in MIR:
1. `NonNull::<T>::eq` does `(a as *mut T) == (b as *mut T)`, but it could just compare the `*const T`s, so this removes `PtrToPtr` casts that are on both sides of a pointer comparison, so long as they're not fat-to-thin casts.
2. `NonNull::<T>::addr` does `transmute::<_, usize>(p as *const ())`, but so long as `T: Thin` that cast doesn't do anything, and thus we can directly transmute the `*const T` instead.
r? mir-opt
It changes behaviour and can cause collisions. E.g. for the following snippet
```rs
mod foo {
pub mod bar {}
pub const bar: i32 = 8;
}
// tranforming the below to `use foo::bar;` causes the error:
//
// the name `bar` is defined multiple times
use foo::bar::{self};
const bar: u32 = 99;
fn main() {
let local_bar = bar;
}
```
we still normalize
```rs
use foo::bar;
use foo::bar::{self};
```
to `use foo::bar;` because this cannot cause collisions.
See: https://github.com/rust-lang/rust-analyzer/pull/17140#issuecomment-2079189725
minor: Bump `actions/download-artifact` and `upload-artifact`
These will stop working in a couple of days. I think we could still use them in a more efficient way, but more tweaking is needed for that.
Various refactorings to rustc_interface
This should make it easier to move the driver interface away from queries in the future. Many custom drivers call queries like `queries.global_ctxt()` before they are supposed to be called, breaking some things like certain `--print` and `-Zunpretty` options, `-Zparse-only` and emitting the dep info at the wrong point in time. They are also not actually necessary at all. Passing around the query output manually would avoid recomputation too and would be just as easy. Removing driver queries would also reduce the amount of global mutable state of the compiler. I'm not removing driver queries in this PR to avoid breaking the aforementioned custom drivers.
transmute size check: properly account for alignment
Fixes another place where ZST alignment was ignored when checking whether something is a newtype. I wonder how many more of these there are...
Fixes https://github.com/rust-lang/rust/issues/101084
Allow constraining opaque types during various unsizing casts
allows unsizing of tuples, arrays and Adts to constraint opaque types in their generic parameters to concrete types on either side of the unsizing cast.
Also allows constraining opaque types during trait object casts that only differ in auto traits or lifetimes.
cc #116652
Save 2 pointers in `TerminatorKind` (96 → 80 bytes)
These things don't need to be `Vec`s; boxed slices are enough.
The frequent one here is call arguments, but MIR building knows the number of arguments from the THIR, so the collect is always getting the allocation right in the first place, and thus this shouldn't ever add the shrink-in-place overhead.
std: refactor the TLS implementation
As discovered by Mara in #110897, our TLS implementation is a total mess. In the past months, I have simplified the actual macros and their expansions, but the majority of the complexity comes from the platform-specific support code needed to create keys and register destructors. In keeping with #117276, I have therefore moved all of the `thread_local_key`/`thread_local_dtor` modules to the `thread_local` module in `sys` and merged them into a new structure, so that future porters of `std` can simply mix-and-match the existing code instead of having to copy the same (bad) implementation everywhere. The new structure should become obvious when looking at `sys/thread_local/mod.rs`.
Unfortunately, the documentation changes associated with the refactoring have made this PR rather large. That said, this contains no functional changes except for two small ones:
* the key-based destructor fallback now, by virtue of sharing the implementation used by macOS and others, stores its list in a `#[thread_local]` static instead of in the key, eliminating one indirection layer and drastically simplifying its code.
* I've switched over ZKVM (tier 3) to use the same implementation as WebAssembly, as the implementation was just a way worse version of that
Please let me know if I can make this easier to review! I know these large PRs aren't optimal, but I couldn't think of any good intermediate steps.
`@rustbot` label +A-thread-locals
minor : Fix duplicate snippets showing up on hover.
With each `config::apply_change` duplicate configs were being added.
Now we first drain the vec that holds these and then start adding. This fixes#17485
internal: Some more small memory optimizations
Not a big impact on metrics, though there are some more savings in queries mainly used by the IDE layer from this
fix: Improve hover text in unlinked file diagnostics
Use full sentences, and mention how to disable the diagnostic if users are intentionally working on unowned files.
(Full disclosure: I've tested a rust-analyzer build in VS Code, but the pop-up logic is currently disabled due to #17062, so I haven't tested that.)
